[quote=Lennart;153183][URL]http://www.us.zyxel.com/Products/Details.aspx?CategoryGroupNo=56E645BA-B447-4DCB-96A0-8B9F732BF4CC[/URL]

/Lennart:smile:[/quote]
Hmm, I see. But...that one looks like just a router with Wi-Fi and HomePlug capability--not a modem. (Possibly that's why the one in Carlos's quote referenced a model number "P-660HWP-D1" rather than just "P-660HWP" as in your link?) :huh:

Reserving the final range of n=596.6K-598K for port 400.

That's all she wrote for files on this drive. :smile:

In the next 2-3 days or so, we'll load up ports 400, 5000, and 4000 with some work for n>600K for k=400-600, 600-800, and 800-1001 respectively so that there is no "break" in the action once the servers finish this drive.

Likely I'll suggest loading each port with n=600K-605K but that's open for discussion. It will be a good speed boost on finding primes, not on the testing itself but on the chance of any particular candidate being a prime; perhaps 3-7%. The sieve depth of P=26T vs. P=5T will help quite a bit.


Gary

[quote=mdettweiler;153230]Hmm, I see. But...that one looks like just a router with Wi-Fi and HomePlug capability--not a modem. (Possibly that's why the one in Carlos's quote referenced a model number "P-660HWP-D1" rather than just "P-660HWP" as in your link?) :huh:[/quote]

O man :google:
[URL]ftp://ftp.zyxel.com/P-660HWP-D1/datasheet/P-660HWP-D1_3.pdf[/URL]

:missingteeth::smile:

/Lennart

[quote=gd_barnes;153237]Reserving the final range of n=596.6K-598K for port 400.

That's all she wrote for files on this drive. :smile:

In the next 2-3 days or so, we'll load up ports 400, 5000, and 4000 with some work for n>600K for k=400-600, 600-800, and 800-1001 respectively so that there is no "break" in the action once the servers finish this drive.

Likely I'll suggest loading each port with n=600K-605K but that's open for discussion. It will be a good speed boost on finding primes, not on the testing itself but on the chance of any particular candidate being a prime; perhaps 3-7%. The sieve depth of P=26T vs. P=5T will help quite a bit.


Gary[/quote]
About how much of >600K to initially load into the servers: I was thinking, maybe we'd be better off loading 600K-610K to start with? Remember, at n=600K, the tests are still going to be almost exactly as big as the ones we're doing right now (so same approximate rate of processing, at least at first). For the 1st Drive, with all 300 k's 400-1001 included, we'd normally load about n=5K at a time into a busy server; but for only 100 k's apiece on each of the three individual ranges, we've got to remember that we'll be progressing 3 times as fast on each one. So...even though our workload will be spread out over all three ranges, I'm thinking that it wouldn't hurt to play it safe and load 600K-610K to start with, at least until we get a feel for how fast we progress on these new divisions of k-ranges.

BTW: on a somewhat unrelated topic, I noticed that we've got two LLRnet 1st Drive ranges that have finished up their last stragglers today, namely 570K-575K for LLRnet G4000, and 582K-587K for IB400. The final results for those ranges should be copied off at the next respective results processing time for both servers. I'll process the ranges tomorrow. :smile:

Max :smile:

[quote=mdettweiler;153251]About how much of >600K to initially load into the servers: I was thinking, maybe we'd be better off loading 600K-610K to start with? Remember, at n=600K, the tests are still going to be almost exactly as big as the ones we're doing right now (so same approximate rate of processing, at least at first). For the 1st Drive, with all 300 k's 400-1001 included, we'd normally load about n=5K at a time into a busy server; but for only 100 k's apiece on each of the three individual ranges, we've got to remember that we'll be progressing 3 times as fast on each one. So...even though our workload will be spread out over all three ranges, I'm thinking that it wouldn't hurt to play it safe and load 600K-610K to start with, at least until we get a feel for how fast we progress on these new divisions of k-ranges.

BTW: on a somewhat unrelated topic, I noticed that we've got two LLRnet 1st Drive ranges that have finished up their last stragglers today, namely 570K-575K for LLRnet G4000, and 582K-587K for IB400. The final results for those ranges should be copied off at the next respective results processing time for both servers. I'll process the ranges tomorrow. :smile:

Max :smile:[/quote]


The reason that you're using for loading in 600K-610K is the reason that I see to load less. I thought about suggesting 600K-603K in each. Had some folks suggested less, I would have gone with it but definitely not more.

Five reasons:

(1) The push will be off and likely some people including myself will pull some of their machines off of the servers.

(2) Since they're spread over 3 k-ranges, people will have more choice causing less processing on each one.

(3) We'll shortly have other efforts going including k=1005-2000 for n=50K-350K, possibly a "mini-drive" for 4-6 k's on the individual-k drive, more of a push on the double-check drive, and a BIG push on sieveing k=1005-2000 for the higher n-ranges so that we can get that started by late Jan.

(4) We want to allow people the option for manual reservations without having the servers suck up a lot of pairs at the lowest n-ranges of the new drives. I suspect that Ian and Carlos will want to reserve some good-sized manual ranges and run fewer on a server than they have been.

(5) We want to gauge demand on each server for the new n-range before loading too much in them. Even though the "new" n=10K range is like a current n=3333 range in a specific server, n=3333 is still a lot of work in our current drive. Just take a look at how long it is taking to finish off port 4000 and that's with the equivalent of ~6 quads on it.

Keep in mind that the project is coming upon a very dynamic stage. We don't know what people will be the most interested in. If we were ONLY doing k=400-1001 for n>600K, that is continuing more with the status quo, then I would definitely agree with you.


Gary

what is against having lots of work in the servers
surely it means less maintenance

[quote=gd_barnes;153257]The reason that you're using for loading in 600K-610K is the reason that I see to load less. I thought about suggesting 600K-603K in each. Had some folks suggested less, I would have gone with it but definitely not more.

Five reasons:

(1) The push will be off and likely some people including myself will pull some of their machines off of the servers.

(2) Since they're spread over 3 k-ranges, people will have more choice causing less processing on each one.

(3) We'll shortly have other efforts going including k=1005-2000 for n=50K-350K, possibly a "mini-drive" for 4-6 k's on the individual-k drive, more of a push on the double-check drive, and a BIG push on sieveing k=1005-2000 for the higher n-ranges so that we can get that started by late Jan.

(4) We want to allow people the option for manual reservations without having the servers suck up a lot of pairs at the lowest n-ranges of the new drives. I suspect that Ian and Carlos will want to reserve some good-sized manual ranges and run fewer on a server than they have been.

(5) We want to gauge demand on each server for the new n-range before loading too much in them. Even though the "new" n=10K range is like a current n=3333 range in a specific server, n=3333 is still a lot of work in our current drive. Just take a look at how long it is taking to finish off port 4000 and that's with the equivalent of ~6 quads on it.

Keep in mind that the project is coming upon a very dynamic stage. We don't know what people will be the most interested in. If we were ONLY doing k=400-1001 for n>600K, that is continuing more with the status quo, then I would definitely agree with you.


Gary[/quote]
Ah, I see what you mean now. Yeah, I guess n=5K would be good after all. :smile:
[quote=henryzz;153277]what is against having lots of work in the servers
surely it means less maintenance[/quote]
As Gary said in his post, primarily so that users wishing to search manual ranges also have a chance to grab some without putting too huge a gap between the server(s)' testing level and that of the manual ranges. As long as we load enough work into each server at a time to last at least 1.5-2 weeks, then that's pretty good as far as maintenance demand goes.

598.0-600.0 done by C443 waiting for Max approval.

LLRnet IB400 has completed 582K-587K.
LLRnet G4000 has completed 570K-575K.

Results emailed to Gary. :smile:

Carlos, I do not have all of the results from 598K-600K. Could you please copy off an updated results file to your website?

[quote=mdettweiler;153386]Carlos, I do not have all of the results from 598K-600K. Could you please copy off an updated results file to your website?[/quote]

Everything is there.